How Good Is Flux Balance Analysis? a Comparison with Nonlinear Simulations of a Simplified Whole-Cell Model

Based on reaction stoichiometries and linear optimization, Flux Balance Analysis (FBA) is widely used to simulate genome-scale metabolism. However, FBA fails to predict some important metabolic phenomena related to physical constraints other than stoichiometry, e.g., the shift from high yield to low yield pathways when nutrients are abundant (overflow metabolism) [1]. To go beyond FBA, other important cellular constraints must be considered, such as the limitation in volume for cell components (e.g. ribosomes, enzymes and metabolites), the limitation in surface area for transporters, the costs of protein production, and the reaction kinetics dependent on enzyme and metabolite concentrations. Fully accounting for those constraints requires computationally inefficient nonlinear optimizations as well as complete knowledge of kinetic parameters, potentially forbidding the simulation of genome-scale models. In order to assess the importance of these constraints in predicting cellular growth and reaction rates, we compared nonlinear simulations of a simplified whole-cell metabolic model (based on previous work [2]) to FBA and to FBA with molecular crowding. FBA growth rate predictions are realistic only at low growth rates, while FBA with molecular crowding provides qualitatively acceptable predictions at all growth rates. However, the flux distributions predicted by both models differ significantly from the nonlinear simulations.

[1] Neijssel & Tempest (1976) Arch Microbiol 110: 305.

[2] Molenaar et al. (2009) Mol Syst Biol 5: 323.